Which is better, Gemini 3.1 Pro or GPT-5.1?

Across 45 shared benchmarks tracked on Vector Wire, Gemini 3.1 Pro leads on 41 and GPT-5.1 leads on 4 — Gemini 3.1 Pro leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 3.1 Pro vs GPT-5.1

GooglevsOpenAI45 shared benchmarks41–4 head-to-head

Benchmark	Gemini 3.1 Pro	GPT-5.1
AA Agentic Index	23	32.2
AA Intelligence	47.7	37.5
AA-LCR	79	76.7
AA-Omniscience	32.9	5.4
arc_agi_1	98	72.8
ARC-AGI-2	77.1	17.6
arena_elo	1486	1439
arena_text_factuality	1471	1445
arena_vision	1280	1250
Artificial Analysis Coding Index	68.8	49.4
browsecomp	85.9	50.8
coding_arena_elo	1447	1421
critpt	17.7	4.9
Fortress	29.8	25.7
frontiermath_tier_4	16.7	12.5
gdpval	23.3	25
GPQA Diamond	94.3	88.1
HLE	51.4	28.5
HLE (with tools)	51.6	42.7
hmmt_nov_2025	94.8	91.7
IFBench	77.1	72.9
LiveCodeBench v6	91.7	87
matharena_visual_math_overall	89.4	76.9
MCP Atlas	78.2	50.1
MMLU-Pro	91	87
mmmlu	92.6	91
MMMU-Pro	83	75.5
multichallenge	71.4	63.4
MultiNRC	64.7	49
OmniScience Accuracy	55.3	37.7
OmniScience Non-Hallucination	50.1	48.1
scicode	59	43.3
simplebench	79.6	53.2
simpleqa_verified	77.3	48.9
SWE-bench Verified	80.6	76.3
TauBench V3 - Banking	21.4	15.9
Terminal-Bench 2.0	68.5	47.6
Terminal-Bench 2.1	74	52.4
Terminal-Bench Hard	68.5	45.5
vectara_answer_rate	99.4	100
vectara_avg_summary_length	107.7	254.4
vectara_factual_consistency	89.6	89.1
vectara_hallucination_rate ↓	10.4	12.1
τ²-Bench Telecom (AA run)	99.3	81.9
τ³-Bench	67.1	14

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.